strate here that p53 uses a palindromic binding site to regulate itstarget gene PAC1. Thus, it is conceivable that p53 may selectivelyregulate different groups of target genes through this mechanism orthrough conventional mechanisms. The identification of thismechanism for p53 action will provide insights into the molecularbasis of how p53 selectively regulates its target genes to eliminatecancer cells and suppress tumorigenesis. A

MethodsCell culture and DNA transfectionEB cells, EB-1 cells and MEFs have been described elsewhere7,18,19. All cancer cell lines werefrom American type culture collection (ATCC). The conditions for cell culture aredescribed in Supplementary Information. LipofectAMINE reagent (Gibco) was used fortransient and stable transfection of cells, according to the manufacturer’s protocol. Forselecting stable clones, transfected cells were grown in medium containing 400 mg ml21

G418 for neomycin resistance, or 2 mg ml21 hygromycin B for hygromycin resistance.

Northern, western and cell cycle analysesTotal RNA was isolated from growing cells using TRIzol reagent (Gibco). Poly(A)þ RNAwas purified using a olyATtract mRNA isolation system (Promega), according to themanufacturer’s instructions. The experimental procedures for northern, western and cellcycle analyses are described in Supplementary Information.

Luciferase reporters, PAC1 siRNA and PAC1 expression plasmidsThe regulatory and promoter region of the human PAC1 gene was cloned by PCR fromnormal human genomic DNA using corresponding primers based on the human genomedatabase. PCR products were subcloned into a pT–Adv vector (Clontech) for sequencingand compared with the genome database. The PAC1 promoter was subcloned into thepGL3-basic reporter (Promega). The deletion of the palindrome in PAC1 of the pGL3-PAC1-711 plasmid was generated using the QuickChange site-directed mutagenesis kit(Stratagene) according to the manufacturer’s protocol. The designs for the constructs,luciferase assay, cellular viability and TdT-mediated dUTP nick end labelling (TUNEL)assay are described in Supplementary Information.

DNase I footprinting analysisThe human PAC1 promoter was end-labelled with [g-32P]ATP by T4 polynucleotidekinase, digested with EcoRVand purified to obtain the sense strand 3 0 -end-labelled probe,and subjected to DNaseI footprinting analysis with the Core Footprinting System(Promega), according to the manufacturer’s instructions. Recombinant p53 protein wasproduced in insect cells infected with a baculovirus vector expressing human wild-typep53, and partially purified through affinity chromatography. The purified recombinantp53 protein was added to bind to radiolabelled probe fragment ð1–2 £ 104 c:p:m:Þ at 37 8Cfor 30 min, followed by the addition of DNase I. Samples were subjected to polyacrylamidegel electrophoresis (PAGE) under denaturing conditions and the dried gel was exposed toautoradiography with an intensifying screen.

Electrophoretic mobility shift assaySynthetic oligonucleotides (pairs of sense and antisense) were annealed and labelled with32P by using T4 polynucleotide kinase and [g-32P]ATP, as described elsewhere20. Briefly,32P-labelled probes were mixed with purified recombinant p53 in a 20-ml DNA bindingreaction buffer. For specificity or competition controls, a labelled random oligonucleotideor excess of unlabelled corresponding oligonucleotide were added together in reactions.For supershift, the anti-p53 monoclonal antibodies (PAb421, PAb1801; Oncogene) wereincluded. The reaction mixtures were incubated at 4 8C for 20 min, resolved by a 4%polyacrylamide gel, and exposed for photography.

Anchorage-independent growth and tumorigenicity assaysExponentially growing cells (6 £ 103 cells per group) were mixed with 3 ml top agarosecontaining 0.35% low melting point agarose in MEM medium with 10% fetal bovineserum (FBS), and seeded onto 3 ml 0.6% solidified agarose in the same medium in six-wellplates. Colonies with a diameter of .1.0 mm were counted after two weeks ofincubation. Results are expressed as means of colony number ^ s.d. of triplicate repeats. Fortumorigenicity assay in vivo, 4-week-old athymic nude mice (BALB/c/nu/nu,Harlan–Sprague–Dawley) were inoculated subcutaneously with 5 £ 105 cells in 0.2 ml MEMmedium. Tumour volume was determined by the equation V ¼ ðL £ W2Þ£ 0:5; where L islength and W is width of tumour. Values are the means ^ s.d. of the counted tumours.

Received 17 December 2002; accepted 20 February 2003; doi:10.1038/nature01519.

1. Levine, A. J. p53, the cellular gatekeeper for growth and division. Cell 88, 323–331 (1997).

2. Vogelstein, B., Lane, D. & Levine, A. J. Surfing the p53 network. Nature 408, 307–310 (2000).

3. Rohan, P. J. et al. PAC-1: a mitogen-induced nuclear protein tyrosine phosphatase. Science 259,

1763–1766 (1993).

4. Ward, Y. et al. Control of MAP kinase activation by the mitogen-induced threonine/tyrosine

phosphatase PAC1. Nature 367, 651–654 (1994).

5. Kastan, M. B. et al. A mammalian cell cycle checkpoint pathway utilizing p53 and GADD45 is defective

in ataxia-telangiectasia. Cell 71, 587–597 (1992).

6. Brugarolas, J. et al. Radiation-induced cell cycle arrest compromised by p21 deficiency. Nature 377,

552–557 (1995).

7. Zhao, R. et al. Analysis of p53-regulated gene expression patterns using oligonucleotide arrays. Genes

Dev. 14, 981–993 (2000).

8. Nishida, E. & Gotoh, Y. The MAP kinase cascade is essential for diverse signal transduction pathways.

Trends Biochem. Sci. 18, 128–131 (1993).

9. Shaw, P. et al. Induction of apoptosis by wild-type p53 in a human colon tumour-derived cell line.

Proc. Natl Acad. Sci. USA 89, 4495–4499 (1992).

10. Yin, Y., Solomon, G., Deng, C. & Barrett, J. C. Differential regulation of p21 by p53 and Rb in cellular

response to oxidative stress. Mol. Carcinog. 24, 15–24 (1999).

11. El Deiry, W. S., Kern, S. E., Pietenpol, J. A., Kinzler, K. W. & Vogelstein, B. Definition of a consensus

binding site for p53. Nature Genet 1, 45–49 (1992).

12. Baker, S. J., Markowitz, S., Fearon, E. R., Willson, J. K. & Vogelstein, B. Suppression of human

colorectal carcinoma cell growth by wild-type p53. Science 249, 912–915 (1990).

13. Foord, O. S., Bhattacharya, P., Reich, Z. & Rotter, V. A DNA binding domain is contained in the

C-terminus of wild type p53 protein. Nucleic Acids Res. 19, 5191–5198 (1991).

14. Gu, W. & Roeder, R. G. Activation of p53 sequence-specific DNA binding by acetylation of the p53

C-terminal domain. Cell 90, 595–606 (1997).

15. El-Deiry, W. S. et al. Topological control of p21WAF1/CIP1 expression in normal and neoplastic

tissues. Cancer Res. 55, 2910–2919 (1995).

16. Fire, A. et al. Potent and specific genetic interference by double-stranded RNA in Caenorhabditis

elegans. Nature 391, 806–811 (1998).

17. Suit, G. et al. A DNA vector-based RNAi technology to suppress gene expression in mammalian cells.

Proc. Natl Acad. Sci. USA 99, 5515–5520 (2002).

18. Almasan, A. et al. Deficiency of retinoblastoma protein leads to inappropriate S-phase entry, activation

of E2F-responsive genes, and apoptosis. Proc. Natl Acad. Sci. USA 92, 5436–5440 (1995).

19. Yin, Y. et al. Involvement of p85 in p53-dependent apoptotic response to oxidative stress. Nature 391,

707–710 (1998).

20. Maiyar, A. C., Huang, A. J., Phu, P. T., Cha, H. H. & Firestone, G. L. p53 stimulates promoter activity of

the sgk. serum/glucocorticoid-inducible serine/threonine protein kinase gene in rodent mammary

epithelial cells. J. Biol. Chem. 271, 12414–12422 (1996).

21. Minden, A. et al. c-Jun N-terminal phosphorylation correlates with activation of the JNK subgroup but

not the ERK subgroup of mitogen-activated protein kinases. Mol. Cell. Biol. 14, 6683–6688 (1994).

Supplementary Information accompanies the paper on Nature’s website

(ç http://www.nature.com/nature).

Acknowledgements We thank A. J. Levine for advice on promoter studies and for providing

critical reagents, and W. Gu for the pCMV-p53D270 plasmid. We are grateful to H. B. Lieberman

for critical reading of the manuscript. We also thank T. A. Sato for help with the luciferase assay.

This work was supported by a start up fund from the Columbia University (Y.Y.) and NIH grants

(to Y.Y. and E.J.H).

Competing interests statement The authors declare that they have no competing financial

interests.

Correspondence and requests for materials should be addressed to Y.Y.

(e-mail: yy151@columbia.edu).

..............................................................

The voltage dependence of NADPHoxidase reveals why phagocytesneed proton channelsThomas E. DeCoursey, Deri Morgan & Vladimir V. Cherny

Department of Molecular Biophysics and Physiology, Rush Presbyterian St Luke’sMedical Center, 1750 West Harrison, Chicago, Illinois 60612, USA.............................................................................................................................................................................

The enzyme NADPH oxidase in phagocytes is important in thebody’s defence against microbes: it produces superoxide anions(O2

2, precursors to bactericidal reactive oxygen species1). Elec-trons move from intracellular NADPH, across a chain compris-ing FAD (flavin adenine dinucleotide) and two haems, to reduceextracellular O2 to O2

2. NADPH oxidase is electrogenic2, generat-ing electron current (Ie) that is measurable under voltage-clampconditions3,4. Here we report the complete current–voltagerelationship of NADPH oxidase, the first such measurement ofa plasma membrane electron transporter. We find that Ie isvoltage-independent from 2100 mV to >0 mV, but is steeplyinhibited by further depolarization, and is abolished at about1190 mV. It was proposed that H1 efflux2 mediated by voltage-gated proton channels5,6 compensates Ie, because Zn21 and Cd21

inhibit both H1 currents7–9 and O22 production10. Here we

show that COS-7 cells transfected with four NADPH oxidase

letters to nature

NATURE | VOL 422 | 3 APRIL 2003 | www.nature.com/nature 531© 2003 Nature Publishing Group

components11, but lacking H1 channels12, produce O22 in the

presence of Zn21 concentrations that inhibit O22 production in

neutrophils and eosinophils. Zn21 does not inhibit NADPHoxidase directly, but through effects on H1 channels. H1 chan-nels optimize NADPH oxidase function by preventing membranedepolarization to inhibitory voltages.

Because it is electrogenic, NADPH oxidase ought to be sensitiveto membrane potential. However, this property has been demon-strated only over a limited voltage range3,13. Furthermore, thehypothesis that voltage-gated proton channels compensate most ofthe charge generated by NADPH oxidase function2,5,6,14 has notbeen tested directly, and recently, Kþ was proposed to have asimilar role15. Although Zn2þ or Cd2þ can inhibit O2

2 pro-duction10,16, they do so at higher concentrations than are requiredto inhibit voltage-gated proton currents5,9,17, raising the possibilitythat these metals directly inhibit NADPH oxidase or affect otherprocesses. Here we establish the importance of Hþ currents, andshow that the voltage dependence of NADPH oxidase optimizes itsperformance.

The current–voltage relationship of Ie generated by NADPHoxidase is shown in Fig. 1. Eosinophils in permeabilized-patchconfiguration4,17 were stimulated with the potent respiratory burstagonist, PMA (phorbol myristate acetate), which activates NADPHoxidase, resulting in inward Ie (ref. 17). We isolated Ie by itssensitivity to the NADPH oxidase inhibitor diphenylene iodiniumchloride (DPI)18. Voltage-gated proton currents were inhibited byaddition of 1.5–6.0 mM ZnCl2 to the bathing solution. These highconcentrations of ZnCl2 (free Zn2þ, 0.5–5.0 mM) did not reduce Ie

noticeably. The mean reduction of Ie by 3 mM Zn2þwas 5.9 ^ 6.5%(mean ^ s.e.m., n ¼ 6), ruling out direct inhibition of NADPHoxidase. Therefore, inhibition of O2

2 production by high Zn2þ

concentrations in previous studies10,19 was not due to direct inhi-bition of NADPH oxidase. We compared currents elicited by voltageramps soon after DPI addition (before inhibition was evident) andlater (Fig. 1a, filled squares and open circles, respectively). Fullinhibition of Ie by DPI takes 1–2 min (ref. 13). The net DPI-sensitive

Ie is plotted in Fig. 1b. The similarity of Ie measured during ramps(open triangles) and at constant voltage (filled circles) confirms thatIe reached steady state at each voltage during the ramps. This resultis expected because the turnover rate of NADPH oxidase is,300 s21 (ref. 20).

Ie–V relationships measured in four cells (of ten studied) areplotted in Fig. 1c to illustrate the variability of the data. Theexpected inhibition of electron efflux by depolarization occursonly at large positive voltages, where the Ie –V relationship becomessteeply voltage dependent. Electron transport was abolished atþ165 to þ220 mV (þ188 ^ 19 mV; mean ^ s.d., n ¼ 10). Thestandard redox potential of the NADPH/NADPþ pair is 2320 mV(ref. 21) and that of O2/O2

2 is 2160 mV (ref. 22). Thus, the nominaldriving force for electron movement across the chain is þ160 mV,although this pseudo-equilibrium potential ought to depend onconcentration23. Depolarization of the phagocyte membrane to theequilibrium potential of the electron transport pathway shuts downNADPH oxidase. The Ie –V relationship was surprisingly indepen-dent of voltage from 2100 mV to an inflection point that occurredbetween 0 mV and þ80 mV. In some cells there was weak voltagedependence at negative voltages (for example, open symbols inFig. 1c), but strong rectification of the Ie –V relationship wasobserved consistently. Evidently, a voltage-independent process israte limiting at voltages ,0 mV, and a voltage-sensitive processbecomes rate determining only during extreme depolarization.Electron transfer between the two intramembrane haem groups ingp91phox (ref. 24) is a likely candidate for this voltage sensitiveprocess.

If no compensatory mechanism existed, Ie in an eosinophil at37 8C would depolarize the membrane by ,1 kV min21 during therespiratory burst, driving the membrane from its resting potentialof 260 mV (ref. 25) to þ190 mV in ,20 ms, and O2

2 productionwould cease. For continuous function of NADPH oxidase, thischarge movement must be balanced by an efficient mechanism thatresponds rapidly to depolarization. Voltage-gated proton channelsare ideally suited to this task, because they are activated by cytosolicacidification and depolarization5,8,17,26. Depolarization to þ20 mVactivates sufficient outward Hþ current to compensate Ie completelyin a PMA-stimulated eosinophil17. However, the only evidencesupporting this role is the inhibition of O2

2 production by Cd2þ

Figure 1 Current–voltage relationship for NADPH-oxidase-mediated electron current (Ie)

in human eosinophils. a, Uncorrected averaged currents recorded in an eosinophil during

voltage ramps early (filled squares) and late (open circles) after addition of DPI to a cell that

was previously stimulated with 60 nM PMA. Voltage ramps were applied at 5-s intervals

from 2100 mV toþ200 mV. The membrane was held at 260 mV after each ramp, and

stepped to 2100 mV for 400 ms before the start of the ramp. To inhibit Hþ channels,

5 mM ZnCl2 was added to all bath solutions. The average of three ramp currents recorded

,4 min after DPI was added was subtracted point-by-point from the average of four

records obtained within 20 s after DPI addition. The difference (b) is DPI-sensitive Ie. Filled

symbols indicate ‘steady-state’ net Ie measured at the end of the 400-ms prepulse to

2100 mV and after 4.4 s at 260 mV. c, Net Ie in four eosinophils representative of the

variability observed in Ie–V relationships in ten cells.

Figure 2 Factors affecting superoxide production. a, Sensitivity of the maximum rate of

superoxide production to ZnCl2 in human eosinophils (EOS, filled triangles), neutrophils

(PMN, filled circles) and COSphox cells (COS, open squares) (mean ^ s.e.m. of 4, 9 or 6

experiments, respectively). Release of O22 at 37 8C was quantified by reduction of

cytochrome c as described previously17, except that cells were suspended in Ringer’s

solution to avoid chelation of Zn2þ. Inhibition was significant at ½ZnCl2�$ 0:3 mM

(P , 0.01, Student’s unpaired t-test). b, Restoration of O22 production (mean and s.e.m.)

by 60 mM CCCP (open symbols), in human eosinophils (EOS, filled triangles, n ¼ 3) or

neutrophils (PMN, filled circles, n ¼ 4) inhibited by Zn2þ. For both cell types, O22

production at 1 or 3 mM ZnCl2 was significantly greater in the presence of CCCP

(P , 0.05). CCCP alone appeared to partially inhibit O22 release, perhaps by uncoupling

mitochondria, although this was not statistically significant. The control data are a subset

of the data in a. All values are normalized to those in the absence of Zn2þ.

letters to nature

NATURE | VOL 422 | 3 APRIL 2003 | www.nature.com/nature532 © 2003 Nature Publishing Group

and Zn2þ (ref. 10) at concentrations larger than those needed toinhibit voltage-gated proton current9,16. If Zn2þ inhibits O2

2 pro-duction by inhibiting Hþ currents, then it should not affect O2

2

production in cells that do not express voltage-gated protonchannels. The only cells known to lack Hþ channels are COS-7cells12. Therefore, we studied O2

2 production in COS-7 cells trans-fected with the four main NADPH oxidase components, gp91phox,p47phox, p67phox and p22phox, which we call COSphox cells11,12. In Fig. 2the concentration–response relationship for ZnCl2 inhibition of themaximal rate of O2

2 production is compared in human eosinophils(filled triangles), neutrophils (filled circles) and COSphox cells (opensquares) stimulated with 60 nM PMA. O2

2 generation was measuredby a standard cytochrome c reduction assay17. There was little effectof up to 3 mM ZnCl2 on COSphox cells, which lack Hþ channels. Incontrast, 0.1–0.3 mM ZnCl2 clearly reduced O2

2 production inhuman phagocytes, with nearly complete inhibition at 3 mMZnCl2. Evidently, ZnCl2 reduces O2

2 production by inhibitingvoltage-gated proton channels, and cannot do so in cells that lackHþ channels. The charge-compensating mechanism of COSphox

cells is not known, but it evidently is less sensitive to Zn2þ than areHþ channels. Figure 2b illustrates that the inhibition of O2

2

production in phagocytes by Zn2þ was partially overcome byaddition of the protonophore, CCCP (carbonyl cyanide m-chlor-ophenylhydrazone). In contrast, the Kþ ionophore, valinomycin,exacerbates the inhibition of O2

2 production by Cd2þ or Zn2þ (refs6, 10). Together, these results indicate that proton efflux rather thanflux of another ion compensates for charge separation by NADPHoxidase in phagocytes.

A small fraction of Ie might be compensated by Kþ, as wasproposed recently15. However, we rule out significant involvementof Kþ current in compensating charge. First, although unstimulatedeosinophils have inwardly rectifying Kþ channels, they have nooutward Kþ conductance, and blocking the inward rectifier doesnot compromise O2

2 production27. In PMA-stimulated eosinophilsin which Ie attests to NADPH oxidase activity, we do not detectoutward Kþ currents (data not shown). Furthermore, upon stimu-lation by PMA, the membrane potential of eosinophils depolarizesto the Nernst potential for Hþ over a wide range of pH gradients28,demonstrating that the predominant conductance is proton selec-tive. Thus, Hþ efflux compensates most of the charge separation byNADPH oxidase in active phagocytes. Although any Kþ effluxwould contribute to charge compensation, the primary role of Kþ

efflux is regulation of the phagosomal volume, ionic strength andpH (ref. 15). Kþ efflux, in contrast to Hþ efflux, causes osmotic andpH changes in the phagosome, which promote activity of proteo-lytic enzymes15. Similarly, complete depletion of phagosomal Cl2 byflux into cytosol would compensate less than 4% of the chargetranslocated by NADPH during the respiratory burst15. Proton fluxis ideally suited to charge compensation, because it is osmoticallyneutral and pH neutral.

Voltage-gated proton channels are highly sensitive to divalentcations7–9, with significant inhibition by 1 mM ZnCl2 at pH $ 7 ineosinophils17. Paradoxically, much higher concentrations arerequired to inhibit O2

2 production. The Ie –V relationship reportedhere explains this apparent discrepancy. Zn2þ and Cd2þ donot ‘block’ Hþ channels by steric occlusion, but rather shift Hþ

current activation to more positive voltages5,9,26. Because Ie isvoltage independent from the normal resting potential to beyond0 mV (Fig. 1), shifting the threshold for Hþ current activationwithin this voltage range does not inhibit NADPH oxidase. Analo-gously, depolarization to ,0 mV with high Kþ concentration hadno effect on O2

2 production by eosinophils27, and only partiallyinhibited O2

2 production by neutrophils29. Distinct inhibition ofO2

2 production in neutrophils first occurred at 100–300 mM Zn2þ

(Fig. 2), which would shift the threshold for activating Hþ channelsat pH 7.4 by 80–90 mV (ref. 9); that is, from 220 mV in activatedphagocytes4 to þ60 to þ70 mV. Consequently, activation of suffi-

cient Hþ efflux to compensate Ie would occur only at voltages withinthe range where Ie is inhibited directly by voltage. In conclusion,Zn2þ inhibits O2

2 production by shifting Hþ channel activationinto or beyond the voltage-dependent region of the Ie –Vrelationship.

During the respiratory burst in neutrophils, the membranedepolarizes to þ58 mV (ref. 30), which is close to the point atwhich depolarization begins to inhibit Ie. However, in spite of thisdepolarization of .100 mV from the resting potential, there isminimal ‘self-inhibition’, because Ie is practically voltage indepen-dent throughout this voltage range. The average reduction of Ie atþ58 mV, relative to Ie at the ‘resting potential’ of 260 mV, was only24 ^ 15% (mean ^ s.d., n ¼ 10). The surprisingly large range overwhich NADPH oxidase is insensitive to membrane potential pro-vides a safety factor that ensures optimal function of this enzymeunless it is confronted by drastic membrane depolarization. Thedepolarization that occurs during the respiratory burst is sufficientto activate compensatory Hþ efflux through Hþ channels withoutsignificantly inhibiting the NADPH oxidase. A

MethodsEosinophil and neutrophil isolationVenous blood was drawn from healthy adult volunteers under informed consent, asapproved by our Institutional Review Board and in accordance with Federal regulations.Neutrophils were isolated by density-gradient centrifugation17, and were suspended in10 mM HEPES-buffered HBSS (with Ca2þ and Mg2þ) at pH 7.4 for O2

2 measurements.Eosinophils were isolated from the neutrophils by negative selection using anti-CD16immunomagnetic beads17. Patch-clamp studies were done on freshly isolated eosinophils,and on eosinophils incubated overnight at 37 8C in RPMI 1640 medium containing25 mM HEPES and L-glutamine (Gibco), supplemented with 10% fetal bovine serum(Bio-Whittaker), 100 units ml21 penicillin, 100 mg ml21 streptomycin (Sigma), and1 ng ml21 recombinant human GM-CSF (R & D Systems). Cells were provided by Larry L.Thomas.

COSphox cellsCOS-7 cells stably transfected with the four main subunits of NADPH oxidase—gp91phox,p22phox, p47phox and p67phox (COSphox cells)—were developed by Price et al.11 and wereprovided by M. Dinauer. COSphox cells were maintained in suspension as described12.

ElectrophysiologyFor permeabilized-patch recording, the pipette solutions contained 80 mM KCH3SO3 or100 mM tetramethylammonium methanesulphonate, 50 mM NH4

þ in the form of 25 mM(NH4)2SO4, 2 mM MgCl2, 5 mM BES, 1 mM EGTA, titrated to pH 7.0, and ,500 mg ml21

solubilized amphotericin B (Sigma). The symmetrical 50 mM NH4þ gradient ‘clamped’ the

intracellular pH to extracellular pH (ref. 4). The bath solution was identical to the100 mM tetramethylammonium methanesulphonate pipette solution, but lackedamphotericin B. Studies were done at 20–25 8C. PMA and CCCP were obtained fromSigma. Other details of electrophysiological measurements are described elsewhere17.

Received 14 October 2002; accepted 25 February 2003; doi:10.1038/nature01523.

1. Babior, B. M. NADPH oxidase: an update. Blood 93, 1464–1476 (1999).

2. Henderson, L. M., Chappell, J. B. & Jones, O. T. The superoxide-generating NADPH oxidase of

human neutrophils is electrogenic and associated with an Hþ channel. Biochem. J. 246, 325–329

(1987).

3. Schrenzel, J. et al. Electron currents generated by the human phagocyte NADPH oxidase. Nature 392,

734–737 (1998).

4. DeCoursey, T. E., Cherny, V. V., Zhou, W. & Thomas, L. L. Simultaneous activation of NADPH

oxidase-related proton and electron currents in human neutrophils. Proc. Natl Acad. Sci. USA 97,

6885–6889 (2000).

5. DeCoursey, T. E. & Cherny, V. V. Potential, pH, and arachidonate gate hydrogen ion currents in

human neutrophils. Biophys. J. 65, 1590–1598 (1993).

6. Henderson, L. M., Chappell, J. B. & Jones, O. T. Internal pH changes associated with the activity of

NADPH oxidase of human neutrophils. Further evidence for the presence of an Hþ conducting

channel. Biochem. J. 251, 563–567 (1988).

7. Mahaut-Smith, M. The effect of zinc on calcium and hydrogen ion currents in intact snail neurones.

J. Exp. Biol. 145, 455–469 (1989).

8. Thomas, R. C. & Meech, R. W. Hydrogen ion currents and intracellular pH in depolarized voltage-

clamped snail neurones. Nature 299, 826–828 (1982).

9. Cherny, V. V. & DeCoursey, T. E. pH-dependent inhibition of voltage-gated Hþ currents in rat alveolar

epithelial cells by Zn2þ and other divalent cations. J. Gen. Physiol. 114, 819–838 (1999).

10. Henderson, L. M., Chappell, J. B. & Jones, O. T. Superoxide generation by the electrogenic NADPH

oxidase of human neutrophils is limited by the movement of a compensating charge. Biochem. J. 255,

285–290 (1988).

11. Price, M. O. et al. Creation of a genetic system for analysis of the phagocyte respiratory burst: high-

level reconstitution of the NADPH oxidase in a nonhematopoietic system. Blood 99, 2653–2661

(2002).

12. Morgan, D., Cherny, V. V., Price, M. O., Dinauer, M. C. & DeCoursey, T. E. Absence of proton channels

in COS-7 cells expressing functional NADPH oxidase components. J. Gen. Physiol. 119, 571–580 (2002).

letters to nature

NATURE | VOL 422 | 3 APRIL 2003 | www.nature.com/nature 533© 2003 Nature Publishing Group

13. Cherny, V. V., Henderson, L. M., Xu, W., Thomas, L. L. & DeCoursey, T. E. Activation of NADPH

oxidase-related proton and electron currents in human eosinophils by arachidonic acid. J. Physiol.

(Lond.) 535, 783–794 (2001).

14. Nanda, A. & Grinstein, S. Protein kinase C activates an Hþ (equivalent) conductance in the plasma

membrane of human neutrophils. Proc. Natl Acad. Sci. USA 88, 10816–10820 (1991).

15. Reeves, E. P. et al. Killing activity of neutrophils is mediated through activation of proteases by Kþ

flux. Nature 416, 291–297 (2002).

16. Kapus, A., Szaszi, K. & Ligeti, E. Phorbol 12-myristate 13-acetate activates an electrogenic Hþ-

conducting pathway in the membrane of neutrophils. Biochem. J. 281, 697–701 (1992).

17. DeCoursey, T. E., Cherny, V. V., DeCoursey, A. G., Xu, W. & Thomas, L. L. Interactions between

NADPH oxidase-related proton and electron currents in human eosinophils. J. Physiol. (Lond.) 535,

767–781 (2001).

18. Cross, A. R. & Jones, O. T. The effect of the inhibitor diphenylene iodonium on the superoxide-

generating system of neutrophils. Specific labelling of a component polypeptide of the oxidase.

Biochem. J. 237, 111–116 (1986).

19. Lowenthal, A. & Levy, R. Essential requirement of cytosolic phospholipase A2 for activation of the Hþ

channel in phagocyte-like cells. J. Biol. Chem. 274, 21603–21608 (1999).

20. Koshkin, V., Lotan, O. & Pick, E. Electron transfer in the superoxide-generating NADPH oxidase

complex reconstituted in vitro. Biochim. Biophys. Acta 1319, 139–146 (1997).

21. Burton, K. & Wilson, T. H. The free-energy changes for the reduction of diphosphopyridine nucleotide

and the dehydrogenation of L-malate and L-glycerol 1-phosphate. Biochem. J. 54, 86–94 (1953).

22. Wood, P. M. The redox potential of the system oxygen-superoxide. FEBS Lett. 44, 22–24 (1974).

23. Lauger, P. Electrogenic Ion Pumps (Sinauer Associates, Sunderland, Massachusetts, 1991).

24. Biberstine-Kinkade, K. J. et al. Heme-ligating histidines in flavocytochrome b 558: identification of

specific histidines in gp91phox. J. Biol. Chem. 276, 31105–31112 (2001).

25. Gordienko, D. V. et al. Voltage-activated proton current in eosinophils from human blood. J. Physiol.

(Lond.) 496, 299–316 (1996).

26. Byerly, L., Meech, R. & Moody, W. Rapidly activating hydrogen ion currents in perfused neurones of

the snail, Lymnaea stagnalis. J. Physiol. (Lond.) 351, 199–216 (1984).

27. Tare, M. et al. Inwardly rectifying whole cell potassium current in human blood eosinophils. J. Physiol.

(Lond.) 506, 303–318 (1998).

28. Banfi, B. et al. A novel Hþ conductance in eosinophils: unique characteristics and absence in chronic

granulomatous disease. J. Exp. Med. 190, 183–194 (1999).

29. Martin, M. A., Nauseef, W. M. & Clark, R. A. Depolarization blunts the oxidative burst of human

neutrophils. Parallel effects of monoclonal antibodies, depolarizing buffers, and glycolytic inhibitors.

J. Immunol. 140, 3928–3935 (1988).

30. Jankowski, A. & Grinstein, S. A noninvasive fluorimetric procedure for measurement of membrane

potential. Quantification of the NADPH oxidase-induced depolarization in activated neutrophils.

J. Biol. Chem. 274, 26098–26104 (1999).

Acknowledgements We thank A. R. Cross and L. L. Thomas for discussions, and T. Iastrebova

and J. Murphy for technical assistance. This work was supported in part by the Heart, Lung and

Blood Institute of the National Institutes of Health (T.E.D.).

Competing interests statement The authors declare that they have no competing financial

interests.

Correspondence and requests for materials should be addressed to T.E.D.

(e-mail: tdecours@rush.edu).

..............................................................

Crystal structure of a transcriptionfactor IIIB core interfaceternary complexZ. Sean Juo*, George A. Kassavetis†, Jimin Wang*, E. Peter Geiduschek†& Paul B. Sigler*‡

* Department of Molecular Biophysics & Biochemistry, Yale University,266 Whitney Avenue, New Haven, Connecticut 06520-8114, USA† Center for Molecular Genetics, University of California, San Diego, 9500 GilmanDrive, La Jolla, California 92093-0634, USA‡ Deceased.............................................................................................................................................................................

Transcription factor IIIB (TFIIIB), consisting of the TATA-bind-ing protein (TBP), TFIIB-related factor (Brf1) and Bdp1, is acentral component in basal and regulated transcription by RNApolymerase III1–4. TFIIIB recruits its polymerase to the promoterand subsequently has an essential role in the formation of theopen initiation complex. The amino-terminal half of Brf1 sharesa high degree of sequence similarity with the polymerase IIgeneral transcription factor TFIIB, but it is the carboxy-terminal

half of Brf1 that contributes most of its binding affinity withTBP5–8. The principal anchoring region is located between resi-dues 435 and 545 of yeast Brf1, comprising its homology domainII. The same region also provides the primary interface forassembling Bdp1 into the TFIIIB complex9. We report here a2.95 A resolution crystal structure of the ternary complex con-taining Brf1 homology domain II, the conserved region of TBPand 19 base pairs of U6 promoter DNA. The structure revealsthe core interface for assembly of TFIIIB and demonstrates howthe loosely packed Brf1 domain achieves remarkable bindingspecificity with the convex and lateral surfaces of TBP.

The crystal structure of the Brf1–TBP–DNA ternary complex isshown in Fig. 1. The revealed portion of Brf1 (residues 437–506)extends in parallel with helices H2 and H2 0 across the convex surfaceof TBP. After two nearly 908 turns, the remaining polypeptidepositions itself along the lateral surface of the first structural repeatof TBP (Fig. 2a, b). This ‘vine-on-a-tree’ conformation is substan-tially different from previously reported polymerase (pol) II tran-scription ternary complexes involving TBP and DNA10. Thestructure also demonstrates that the convex surface of TBP isextensively engaged as the principal interface for binding. A totalof 31 Brf1 residues in the crystal structure interact with 29 TBPresidues through 31 hydrogen bonds and 114 van der Waalscontacts. The large number of interactions is exceptional, approxi-mately three to five times greater than for the comparable pol IIternary complex structures. As a result of all these contacts, 3,230 A2

of TBP surface area becomes inaccessible to solvent on Brf1 binding.The conformation of Brf1 homology domain II revealed in the

crystal structure is unusually extended, which does not fit thegeneral idea of a structural domain. Given that this Brf1 segment,when unbound, is resistant to protease digestion (Z.S.J., unpub-lished results), it must adopt a stable structural fold in solution andundergo global unfolding on complex formation with TBP. Thisresembles the maintenance of a partially unfolded state of theGTPase-activating protein SptP by its cognate chaperone SicP, asthe latter stabilizes the elongated helical conformation of the targetprotein11. The revealed Brf1 domain does not involve DNA bindingin the crystal structure. Notably, the unresolved segment (residues507–596) contains a proposed cryptic DNA-binding domain12. TheH25 helix of Brf1 lies close to DNA and its trajectory points towardsthe start site of transcription, which is consistent with the DNA-binding potential of this segment.

The conformation of DNA-bound TBP remains similar to therelated pol II ternary complex structures10, with an average root-mean-square (r.m.s.) deviation of 0.54 A for Ca positions despitevariations of sequence. In comparison with the apo-structure13, TBPundergoes noticeable conformational changes on DNA binding(averaged r.m.s. deviation of 1.19 A for Ca positions). The twodirect repeats move slightly closer to accommodate the widenednarrow groove of the TATA element, and an approximately 3.3 Adisplacement is observed at the first stirrup region of TBP whenits second structural repeat is aligned. Owing to the pseudo-symmetrical nature of the yeast U6 TATA motif, its sequence forcrystallization was modified to favour unidirectional binding ofTBP14 (Fig. 2c). As observed previously, the TATA element issmoothly curved and doubly kinked at the termini. The minorgroove width of the downstream B-form DNA is substantiallynarrower, which is a consequence of being predominantly(A þ T)-rich. It is worth noting that the terminal A-tract segmentinteracts firmly with its non-crystallographic symmetry (NCS)counterpart (see Supplementary Information). Given that 20 basepairs of DNA stack between two complexes and maintain a stableconformation without any external support, the stiffness of A-tractsequence must contribute significantly to maintaining local struc-tural stability in the crystal lattice14.

The Brf1 homology domain II is mainly helical. The existence offour helices (H21, H23, H24 and H25) is consistent with secondary

letters to nature

NATURE | VOL 422 | 3 APRIL 2003 | www.nature.com/nature534 © 2003 Nature Publishing Group